Ogaday / demo-app

Single sourcing a python application in a docker container

Single Sourcing Poetry Apps on Docker

Demo of a simple, containerised Python app packaged with Poetry, versioned by git tags.

Poetry is a fantastic tool for packaging python libraries and applications. The poetry-dynamic-versioning plugin thinly wraps the poetry build system and looks up the package version from your source control (such as git tags), and injects them into the build info (ie. package metadata) and optionally into the distributed source code itself, for instance overwriting the __version__ variable in __init__.py.

The advantage of this is that your package version has a single source, tied to version control, and there's no need for multi-step tag increments, and not chance of version strings falling out of sync.

Unfortunately, the docker build process can can cause some complications. Typically, the git repo data is no available during the build process. This project demonstrates how to get around this with a number of strategies. All solutions take advantage of new RUN --mount Dockerfile syntax made available by the BuildKit backend.

Strategies

An additional complexity is that poetry-dynamic-versioning only modifies the version number in the code during the poetry build process, and reverts changes once the command completes. Additionally, the poetry install installs the package in editable mode, and this doesn't appear to be configurable. Therefore, the __version__ variable in the source code remains as the placeholder value. There are a couple of approaches to handle this:

• Use poetry to build a wheel file, and install the wheel in the image using pip. The wheel reflects the dynamic version set by the tool. This strategy is implemented in the main branch.
• Use the poetry-dynamic-versioning command in a docker build step to write the version to the source files. This command overwrites the __version__ placeholder with the dynamic version. (This is also configurable) This strategy is implemented in the dynamic-version-script branch.
• Look up the __version__ using importlib.metadata. This strategy is implemented in the importlib-metadata branch.

There are different trade-offs to consider: For instance, packaging the code as a wheel is the most steps, but might be easier to use in a multistage build.

Requirements

• docker/dockerfile:1.2 to enable the require BuildKit features.

Running

The build process:

docker build -t demo-app .

Running the app container:

docker run --rm -it -p 5000:5000 demo-app

Checking the app version:

curl 127.0.0.1:5000
# {"__version__": "0.1.0"}